New types of edge detectors optimized for noisy conditions are introduced and analyzed. The detectors contain linear masks followed by the median operation. The linear masks are designed for detecting actual edges and not for detecting false edges. Extensive numerical simulations show that these edge detectors give better edge signals, attenuate noise better, and yet give as sharp edges as commonly used

Sobel and Prewitt detectors. The detectors can be easily implemented with common image-processing equipment. Examples on actual images confirm the analytical and simulated results.